Update app.py

app.py

@@ -1,235 +1,63 @@
 import os
-import re
-import base64
-import traceback
-import requests
 import gradio as gr
+import requests
+import inspect
 import pandas as pd
-from google import genai
-from google.genai import types
+import google.generativeai as genai
 
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+GEMINI_MODEL = "gemini-1.5-flash"  # Fast, cost-effective model
 
-
-# ── Utilities ─────────────────────────────────────────────────────────
-
-def clean_answer(text: str) -> str:
-    """Normalize model output for exact-match scoring."""
-    if not text:
-        return ""
-    text = text.strip()
-    # Remove markdown fences
-    text = re.sub(r"^```[a-zA-Z]*\s*", "", text)
-    text = re.sub(r"\s*```$", "", text)
-    # Remove common label prefixes
-    for prefix in [
-        "final answer:", "answer:", "the answer is:",
-        "the final answer is:", "result:", "output:",
-    ]:
-        if text.lower().startswith(prefix):
-            text = text[len(prefix):].strip()
-            break
-    # Collapse whitespace, cap length
-    return " ".join(text.split())[:300]
-
-
-def extract_text(response) -> str:
-    """
-    Extract the final answer text from a GenerateContentResponse.
-    Works for plain text, google_search grounding, and code_execution results.
-    Always returns the LAST meaningful text part (= answer after tool use).
-    """
-    parts_text = []
-    try:
-        for candidate in response.candidates:
-            for part in candidate.content.parts:
-                if hasattr(part, "text") and part.text and part.text.strip():
-                    parts_text.append(part.text.strip())
-                if hasattr(part, "code_execution_result") and part.code_execution_result:
-                    out = getattr(part.code_execution_result, "output", "")
-                    if out and str(out).strip():
-                        parts_text.append(str(out).strip())
-    except Exception:
-        pass
-    if parts_text:
-        return parts_text[-1]
-    try:
-        return (response.text or "").strip()
-    except Exception:
-        return ""
-
-
-def fetch_task_file(task_id: str) -> tuple:
-    """Download file attached to a GAIA task. Returns (bytes, filename)."""
-    try:
-        r = requests.get(f"{DEFAULT_API_URL}/files/{task_id}", timeout=20)
-        if r.status_code == 200 and r.content:
-            cd   = r.headers.get("Content-Disposition", "")
-            name = ""
-            if "filename=" in cd:
-                name = cd.split("filename=")[-1].strip().strip('"')
-            name = name or f"file_{task_id}"
-            print(f"  [file] {name}  ({len(r.content)} bytes)")
-            return r.content, name
-    except Exception as e:
-        print(f"  [file] fetch error: {e}")
-    return None, ""
-
-
-def build_contents(question: str, file_bytes, fname: str) -> list:
-    """Package question + optional file into Gemini contents list."""
-    if file_bytes is None:
-        return [question]
-    ext = fname.rsplit(".", 1)[-1].lower() if "." in fname else ""
-
-    IMAGE_MIME = {"png":"image/png","jpg":"image/jpeg","jpeg":"image/jpeg",
-                  "gif":"image/gif","webp":"image/webp","bmp":"image/bmp"}
-    AUDIO_MIME = {"mp3":"audio/mpeg","wav":"audio/wav","ogg":"audio/ogg",
-                  "flac":"audio/flac","m4a":"audio/mp4"}
-
-    if ext in IMAGE_MIME:
-        return [types.Part.from_bytes(data=file_bytes, mime_type=IMAGE_MIME[ext]), question]
-    if ext == "pdf":
-        return [types.Part.from_bytes(data=file_bytes, mime_type="application/pdf"), question]
-    if ext in AUDIO_MIME:
-        return [types.Part.from_bytes(data=file_bytes, mime_type=AUDIO_MIME[ext]), question]
-
-    # Text-based files: embed as context
-    try:
-        txt = file_bytes.decode("utf-8", errors="replace")
-        return [f"Attached file ({fname}):\n```\n{txt[:12000]}\n```\n\n{question}"]
-    except Exception:
-        b64 = base64.b64encode(file_bytes).decode()
-        return [f"Attached file ({fname}) base64:\n{b64[:2000]}\n\n{question}"]
-
-
-# ── Agent ─────────────────────────────────────────────────────────────
-
-class BasicAgent:
-    """
-    Gemini 2.0 Flash agent with Google Search grounding, code execution,
-    and file-attachment support for GAIA benchmark questions.
-    """
-
-    SYSTEM = """You are a precise expert assistant solving GAIA benchmark evaluation questions.
-
-CRITICAL OUTPUT RULE:
-Output ONLY the final answer — nothing else.
-No explanation, no reasoning, no preamble, no trailing sentence.
-
-FORMAT RULES:
-- Numbers : digits only, no thousand-separators. Drop .0 from whole numbers.
-- Lists   : comma-separated values, alphabetical order unless otherwise specified.
-- Yes/No  : exactly "yes" or "no" (lowercase).
-- Names   : full name unless only first or last is requested.
-- Dates   : match the format implied by the question.
-- Units   : include units only if the question asks for them.
-
-STRATEGY:
-1. Read the question (and any attached file) carefully.
-2. Use Google Search for any fact, date, count, name, or external data you need.
-3. Use code execution for arithmetic, unit conversion, or data analysis.
-4. Think step-by-step internally.
-5. Output ONLY the single final answer."""
-
-    CODE_KEYWORDS = {
-        "calculate","compute","sum","total","average","mean","median",
-        "percentage","multiply","divide","convert","how many","count",
-        "square root","power","factorial","modulo","remainder",
-    }
-
+# --- Gemini Agent Definition ---
+class GeminiAgent:
     def __init__(self):
         api_key = os.getenv("GEMINI_API_KEY")
         if not api_key:
-            raise EnvironmentError("GEMINI_API_KEY environment variable is not set.")
-
-        self.client   = genai.Client(api_key=api_key)
-        self.model    = "gemini-2.0-flash"
-
-        cfg = lambda tools: types.GenerateContentConfig(
-            system_instruction=self.SYSTEM,
-            tools=tools,
-            temperature=0,
-        )
-
-        self.search_cfg = cfg([types.Tool(google_search=types.GoogleSearch())])
-        self.code_cfg   = cfg([types.Tool(code_execution=types.ToolCodeExecution())])
-        self.plain_cfg  = cfg([])   # for file questions (model reads the file itself)
-
-        print(f"BasicAgent initialized (model={self.model})")
+            raise ValueError("GEMINI_API_KEY environment variable not set. Please set it before running.")
+        genai.configure(api_key=api_key)
+        self.model = genai.GenerativeModel(GEMINI_MODEL)
+        print(f"GeminiAgent initialized with model {GEMINI_MODEL}.")
 
     def __call__(self, question: str) -> str:
-        # task_id is embedded in the question only when called from run_and_submit_all;
-        # we extract it via a side-channel attribute set before each call.
-        task_id = getattr(self, "_current_task_id", "")
-        try:
-            return self._run(question, task_id)
-        except Exception:
-            print(traceback.format_exc())
-            # Last-resort plain call — never return empty
-            try:
-                r = self.client.models.generate_content(
-                    model=self.model, contents=question, config=self.plain_cfg)
-                ans = clean_answer(extract_text(r))
-                return ans if ans else self._forced_answer(question)
-            except Exception:
-                return self._forced_answer(question)
-
-    def _run(self, question: str, task_id: str) -> str:
-        file_bytes, fname = fetch_task_file(task_id) if task_id else (None, "")
-        contents          = build_contents(question, file_bytes, fname)
-        q_lower           = question.lower()
-
-        has_file   = file_bytes is not None
-        needs_code = not has_file and any(kw in q_lower for kw in self.CODE_KEYWORDS)
-
-        config = self.plain_cfg if has_file else (
-                 self.code_cfg  if needs_code else
-                 self.search_cfg)
-
-        resp = self.client.models.generate_content(
-            model=self.model, contents=contents, config=config)
-        raw  = extract_text(resp)
-        ans  = clean_answer(raw)
-        print(f"  raw : {raw[:120]!r}")
-        print(f"  ans : {ans!r}")
-
-        # If empty, retry with search
-        if not ans:
-            resp2 = self.client.models.generate_content(
-                model=self.model, contents=contents, config=self.search_cfg)
-            ans = clean_answer(extract_text(resp2))
-            print(f"  retry: {ans!r}")
-
-        # Still empty → force a plain answer (never return blank)
-        if not ans:
-            ans = self._forced_answer(question)
-
-        return ans
-
-    def _forced_answer(self, question: str) -> str:
-        """Absolute last resort — plain call with no tools, no format rules."""
+        print(f"GeminiAgent received question (first 50 chars): {question[:50]}...")
+        # Prompt engineering: ask for a direct, accurate answer. Avoid "I don't know" or similar vague responses.
+        prompt = f"""
+You are an expert assistant. Answer the following question concisely and accurately.
+Do not use phrases like "I don't know", "NA", "not applicable", or leave the answer empty.
+If the question is multiple choice, give the letter of the correct answer followed by the answer text.
+If the question asks for a number, give the number only.
+If the question asks for a list, provide it clearly.
+
+Question: {question}
+
+Answer:
+"""
         try:
-            r = self.client.models.generate_content(
-                model=self.model,
-                contents=f"Answer in one word or number only:\n{question}",
-                config=types.GenerateContentConfig(temperature=0),
-            )
-            ans = clean_answer(extract_text(r))
-            return ans if ans else "unknown"
-        except Exception:
-            return "unknown"
-
-
-# ── Gradio runner (original HF template structure preserved) ──────────
+            response = self.model.generate_content(prompt)
+            answer = response.text.strip()
+            # Ensure we never return an empty or placeholder answer
+            if not answer or answer.lower() in ["na", "n/a", "i don't know", "unknown"]:
+                # Fallback: try a more specific prompt to force an answer
+                fallback_prompt = f"Answer this question directly, without any hedging: {question}"
+                fallback_response = self.model.generate_content(fallback_prompt)
+                answer = fallback_response.text.strip()
+                if not answer:
+                    answer = "The answer could not be determined, but a reasonable response is required."
+            print(f"GeminiAgent returning answer (first 50 chars): {answer[:50]}...")
+            return answer
+        except Exception as e:
+            print(f"Gemini API error: {e}")
+            # Last resort: return a non-empty, generic answer (should rarely happen)
+            return f"I encountered an error, but based on the question '{question[:100]}', a plausible answer is: please consult official sources."
 
 def run_and_submit_all(profile: gr.OAuthProfile | None):
     """
-    Fetches all questions, runs the BasicAgent on them, submits all answers,
+    Fetches all questions, runs the GeminiAgent on them, submits all answers,
     and displays the results.
     """
-    space_id = os.getenv("SPACE_ID")
+    # --- Determine HF Space Runtime URL and Repo URL ---
+    space_id = os.getenv("SPACE_ID")  # Get the SPACE_ID for sending link to the code
 
     if profile:
         username = f"{profile.username}"
@@ -238,18 +66,19 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         print("User not logged in.")
         return "Please Login to Hugging Face with the button.", None
 
-    api_url       = DEFAULT_API_URL
+    api_url = DEFAULT_API_URL
     questions_url = f"{api_url}/questions"
-    submit_url    = f"{api_url}/submit"
+    submit_url = f"{api_url}/submit"
 
-    # 1. Instantiate Agent
+    # 1. Instantiate Agent (modified to GeminiAgent)
     try:
-        agent = BasicAgent()
+        agent = GeminiAgent()
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
 
-    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
+    # In the case of an app running as a Hugging Face space, this link points toward your codebase
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else "https://github.com/your-repo"
     print(agent_code)
 
     # 2. Fetch Questions
@@ -259,54 +88,56 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         response.raise_for_status()
         questions_data = response.json()
         if not questions_data:
+            print("Fetched questions list is empty.")
             return "Fetched questions list is empty or invalid format.", None
         print(f"Fetched {len(questions_data)} questions.")
     except requests.exceptions.RequestException as e:
+        print(f"Error fetching questions: {e}")
         return f"Error fetching questions: {e}", None
+    except requests.exceptions.JSONDecodeError as e:
+        print(f"Error decoding JSON response from questions endpoint: {e}")
+        print(f"Response text: {response.text[:500]}")
+        return f"Error decoding server response for questions: {e}", None
     except Exception as e:
+        print(f"An unexpected error occurred fetching questions: {e}")
         return f"An unexpected error occurred fetching questions: {e}", None
 
-    # 3. Run Agent
-    results_log     = []
+    # 3. Run your Agent
+    results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
-
     for item in questions_data:
-        task_id       = item.get("task_id")
+        task_id = item.get("task_id")
         question_text = item.get("question")
         if not task_id or question_text is None:
             print(f"Skipping item with missing task_id or question: {item}")
             continue
-
-        print(f"\n[{task_id}]  {question_text[:120]}")
-        agent._current_task_id = task_id   # pass task_id for file download
-
         try:
             submitted_answer = agent(question_text)
+            answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
         except Exception as e:
-            submitted_answer = "unknown"
             print(f"Error running agent on task {task_id}: {e}")
-
-        print(f"  → submitted: {submitted_answer!r}")
-        answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
-        results_log.append({
-            "Task ID":          task_id,
-            "Question":         question_text,
-            "Submitted Answer": submitted_answer,
-        })
+            # Provide a non-empty fallback
+            fallback = f"An error occurred, but a reasonable answer to '{question_text[:100]}' is: check official documentation."
+            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": fallback})
+            answers_payload.append({"task_id": task_id, "submitted_answer": fallback})
 
     if not answers_payload:
+        print("Agent did not produce any answers to submit.")
         return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
 
     # 4. Prepare Submission
     submission_data = {
-        "username":   username.strip(),
+        "username": username.strip(),
         "agent_code": agent_code,
-        "answers":    answers_payload,
+        "answers": answers_payload
     }
-    print(f"Submitting {len(answers_payload)} answers for user '{username}'...")
+    status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
+    print(status_update)
 
     # 5. Submit
+    print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
     try:
         response = requests.post(submit_url, json=submission_data, timeout=60)
         response.raise_for_status()
@@ -319,7 +150,8 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             f"Message: {result_data.get('message', 'No message received.')}"
         )
         print("Submission successful.")
-        return final_status, pd.DataFrame(results_log)
+        results_df = pd.DataFrame(results_log)
+        return final_status, results_df
     except requests.exceptions.HTTPError as e:
         error_detail = f"Server responded with status {e.response.status_code}."
         try:
@@ -329,30 +161,37 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             error_detail += f" Response: {e.response.text[:500]}"
         status_message = f"Submission Failed: {error_detail}"
         print(status_message)
-        return status_message, pd.DataFrame(results_log)
+        results_df = pd.DataFrame(results_log)
+        return status_message, results_df
     except requests.exceptions.Timeout:
-        return "Submission Failed: The request timed out.", pd.DataFrame(results_log)
+        status_message = "Submission Failed: The request timed out."
+        print(status_message)
+        results_df = pd.DataFrame(results_log)
+        return status_message, results_df
     except requests.exceptions.RequestException as e:
-        return f"Submission Failed: Network error - {e}", pd.DataFrame(results_log)
+        status_message = f"Submission Failed: Network error - {e}"
+        print(status_message)
+        results_df = pd.DataFrame(results_log)
+        return status_message, results_df
     except Exception as e:
-        return f"An unexpected error occurred during submission: {e}", pd.DataFrame(results_log)
-
+        status_message = f"An unexpected error occurred during submission: {e}"
+        print(status_message)
+        results_df = pd.DataFrame(results_log)
+        return status_message, results_df
 
 # --- Build Gradio Interface using Blocks ---
 with gr.Blocks() as demo:
-    gr.Markdown("# Basic Agent Evaluation Runner")
+    gr.Markdown("# Gemini Agent Evaluation Runner")
     gr.Markdown(
         """
         **Instructions:**
-
-        1.  Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
+        1.  Clone this space, then modify the code to adjust prompts or model parameters as needed.
         2.  Log in to your Hugging Face account using the button below. This uses your HF username for submission.
-        3.  Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
-
+        3.  Click 'Run Evaluation & Submit All Answers' to fetch questions, run the Gemini agent, submit answers, and see the score.
         ---
         **Disclaimers:**
-        Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).
-        This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.
+        Submitting may take some time (agent processes all questions sequentially).  
+        This setup uses Google's Gemini API – ensure `GEMINI_API_KEY` is set as a secret in your Space or environment.
         """
     )
 
@@ -370,8 +209,10 @@ with gr.Blocks() as demo:
 
 if __name__ == "__main__":
     print("\n" + "-"*30 + " App Starting " + "-"*30)
+    # Check for environment variables
     space_host_startup = os.getenv("SPACE_HOST")
-    space_id_startup   = os.getenv("SPACE_ID")
+    space_id_startup = os.getenv("SPACE_ID")
+    gemini_key = os.getenv("GEMINI_API_KEY")
 
     if space_host_startup:
         print(f"✅ SPACE_HOST found: {space_host_startup}")
@@ -386,6 +227,12 @@ if __name__ == "__main__":
     else:
         print("ℹ️  SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
 
+    if gemini_key:
+        print("✅ GEMINI_API_KEY is set.")
+    else:
+        print("⚠️  WARNING: GEMINI_API_KEY environment variable is not set. The agent will fail to initialize.")
+
     print("-"*(60 + len(" App Starting ")) + "\n")
-    print("Launching Gradio Interface for Basic Agent Evaluation...")
+
+    print("Launching Gradio Interface for Gemini Agent Evaluation...")
     demo.launch(debug=True, share=False)